Control of a ventilator machine

Abstract

 The present invention relates to the automatic control of a ventilator machine, for changing over between two alternating phases of ventilation by, in one phase of ventilation, causing a control unit to examine a sensed respiratory signal for breathing activity (μV) for a threshold criterion (2) for the changeover to the next phase of ventilation. There is in an expiration phase, for the changeover to an inspiration phase, a dynamic threshold curve, and there is in an inspiration phase a dynamic threshold curve, and a changeover is made when the signal for breathing activity threshold curve is crossed. The durations of the inspiration and expiration phases or the duration of a breath (an inspiration and an expiration phase) are each stored and the expected times of the maximum phase durations are derived from the distributions of the phase durations.

Description

[0001] The present invention relates to ventilator machine and a method for the automatic control of a ventilator machine, for changing over between two alternating phases of ventilation (inspiration and expiration) by, in one phase of ventilation, causing a control unit to examine a sensed respiratory signal for breathing activity for a threshold criterion for the changeover to the next phase of ventilation and changing over from one phase of ventilation to the other when the threshold criterion is met.

[0002] The aim of the artificial maintaining of respiration with ventilator machines is to relieve the strain on a patient's respiratory muscles and to ensure that there is an adequate supply of oxygen and that carbon dioxide is eliminated to an adequate degree. This can be done by causing the ventilator machine to assume responsibility for the whole of the breathing activity or, in assisting techniques, for part of the breathing activity, existing breathing activity by the patient being assisted or boosted in these latter assisting techniques. For this purpose, the ventilator machines contain a ventilator unit to supply gas for breathing at a pressure which is preset by a control unit. Also present are sensors which sense, as a function of time, pneumatic breathing signals, such for example as the airway pressure, or the flow of the gas for breathing and its volume (which is obtained by integrating the flow), and pass these signals on to the control unit.

[0003] In view of the increase in chronic lung disease and the demand for an improved therapy, the non-invasive assistance of breathing with improved interaction between the patient and the ventilator unit is a crucial requirement which needs to be met by modern-day ventilator machines. A significant object to be achieved in this case is the establishing of temporal synchrony between the assistance provided by the machine and the patient's own breathing activity. In the past, what was often done was to sedate patients who were breathing spontaneously to allow the ventilation to be correctly set and to allow synchrony to be forced to exist between the patient and the ventilator machine. From the knowledge which now exists, this procedure is no longer acceptable because risks have to be run of the lungs being damaged by the ventilation.

[0004] For improved synchronisation between the patient's breathing activity and the action of the ventilator unit, it is important for the beginning of inspiration and the beginning of expiration to be reliably detected in the patient's breathing activity at an early point in time. Especially in the case of neonates and COPD patients, the detection of phases of breathing is often incorrect or late with conventional methods and results in increased breathing effort even going as far as exhaustion.

[0005] For artificial maintaining of respiration which is intended to take account of the patient's breathing activity in an improved way, it is known from DE 10 2007 062 214 B3 not only for pneumatic signals for breathing activity to be sensed but also for electromyographic signals to be picked up by electrodes on the thorax and for electromyographic signals for breathing activity (EMG signals) to be derived from these. These EMG signals are independent of the pneumatic signals for breathing activity and thus constitute an independent source of information which can be used to sense the beginning of inspiration and expiration. However it is not uncommon for the EMG signals to have interference and disruptions, such for example as the ECG signal from the heart, motion artefacts, or what is referred to as crosstalk (muscle activity which has nothing to do with the patient's respiratory system), superimposed on them.

[0006] Triggering of breaths on the basis of EMG signals is described in US 6,588,423 B1. In this case the raw EMG signal is pre-processed and, for triggering, a measure of the intensity of the EMG signal (the root mean square) is finally checked, a threshold which is fixed, in relation to one breath, being used.

[0007] In practice however even the pre-processed EMG signal is more susceptible to interference or disruption than pneumatic signals (pressure or flow). A susceptibility of this kind to interference or disruption, or a volatility, makes it more difficult to change over between breaths or trigger them when trigger thresholds are used, because the incorrect triggering of too many breaths may occur (what is referred to as auto-triggering), or breaths may be triggered too late (what is referred to as delayed or missed triggering).

[0008] Although signals can be prevented from being affected by interference or disruptions by suitable filtering (e.g. by means of the formation of sliding means), for the purpose of making a changeover between phase of ventilation this has the serious disadvantage of causing an additional delay to the signals.

[0009] In DE 102 12 497 A1 it is pointed out as a general comment that, at the beginning of an inspiration phase, it is considerably more likely for the inspiration phase to continue than it is at the premature end of the phase and that there is a greater likelihood of a fresh inspiration phase beginning shortly before the end of the expiration phase. The basic rule is said to be that, as it becomes increasingly likely that the event which triggers the phase of ventilation will occur, the threshold for triggering can be lowered, on the one hand because it is becoming less likely that interference or disruptions will have an effect and, what is more, because even if mis-triggering happens due to the occurrence of interference the results of this mis-triggering will be much less of a nuisance due to the closeness in time to a correct changeover time than a mis-triggering at a complete incorrect point in time. However, apart from this no other details are given of how and with what curve over time up to what target point in time a dynamic variation in threshold value over time may be implemented.

[0010] It is an object of the present invention to specify a method for the automatic control of a ventilator machine which on the one hand makes possible a sensitive changeover to the next phase of ventilation and on the other hand keeps incorrect changeovers from inspiration to expiration or from expiration to inspiration as low as possible, by using a dynamic threshold which is well matched to the ventilation situation at the time.

[0011] What is used to achieve this object is a method which has the features given in claim 1. Advantageous embodiments of the invention are specified in the dependent claims.

[0012] In accordance with the invention, there is used in an expiration phase, for the changeover to an inspiration phase, a dynamic threshold curve which, after the beginning of the current expiration phase keeps the threshold, for a selected inspiratory refractory period, i.e. until a time t_i1, at values which are so high that a changeover to inspiration is impossible at such an early time. After this, the threshold curve is lowered in a monotonic decrease to an inspiratory target threshold value at the expected maximum duration t_i2 of the phase, and a changeover is made to the inspiration phase as soon as the signal for breathing activity rises above the threshold curve for inspiration.

[0013] In an inspiration phase, there is used for the changeover to an expiration phase a dynamic threshold curve which, after the beginning of the current inspiration phase, is held for a selected, short expiratory refractory period, i.e. until a time t_e1, at values which are so low that a changeover to expiration is impossible in this early phase. After this the threshold curve is raised in a monotonic increase to an expiratory target threshold value at the expected maximum duration t_e2 of the phase, and a changeover is made to the expiration phase as soon as the signal for breathing activity drops between the threshold curve over the expiration.

[0014] In this connection, the durations of the inspiration and expiration phases or the duration of a breath (the sum of the inspiration phase and expiration phase for one breath) are each stored. The expected maximum phase durations t_i2 and t_e2 can be derived from the distributions of the phase durations (when the reference point is the beginning of the given phase of ventilation) or from the distribution of the breath durations (when the reference point is the beginning of the previous phase of ventilation), preferably in the form of a p quantile of the distribution, the parameter P being fixed in advance and being of a high value close to 1, e.g. 0.95, which means that the time of the expected maximum phase duration is so positioned that in the case of 95% of the previous phases of ventilation the actual end of the phase had already been reached at this time. Alternatively, there may also be assumed to be a Gaussian distribution and the expected maximum phase duration may be fixed as a preset number of standard deviations above the mean in this distribution, e.g. 2.5 σ.

[0015] The times t_i1 and t_e1 of the ends of the inspiratory and expiratory refractory periods and the expected maximum phase durations t_i2 and t_e2 may be referred to the changeover to the current phase of ventilation as a time zero. Alternatively, the waveform of the signal for breathing activity may be stored for at least the duration of one phase of ventilation and the beginning of the current phase of ventilation may be determined retrospectively by examining the waveform of the signal for breathing activity in a period around the changeover to the current phase of ventilation. As a result of a more exact examination of the waveform of the signal for breathing activity, the actual beginning of the current phase of ventilation can be determined more accurately in retrospect than the triggering time for the changeover of the ventilation machine can in real time. The distributions of the phase durations also give the median values t_im and t_em, in the form of 0.5 quantiles, and these can be considered expected values for the durations of the inspiration and expiration phases.

[0016] The target threshold value is determined from the amplitude distributions of the signals for breathing activity at the times of the median phase durations t_im and t_em, t_im being the median value of the durations of the inspiratory phases and t_em being that of the durations of the expiratory phases. In the amplitude distributions at these times, the target threshold values can be fixed as p quantiles or, if a Gaussian distribution is assumed, as a multiple (not generally a whole-number one) of the standard deviation relative to the mean. The inspiratory target threshold value may for example be fixed as the 0.05 quantile in the amplitude distribution at the time t_im, i.e. the target threshold value is so positioned that 95% of the amplitudes of the signal for breathing activity are above the target threshold value at the time t_im. The expiratory target threshold value may be fixed as the 0.95 quantile in the amplitude distribution at the time t_em, i.e. the target threshold value is so positioned that 95% of the amplitudes of the signal for breathing activity are below the expiratory target threshold value at the time t_em.

[0017] In a preferred embodiment of the method, the values of the signal for breathing activity are stored at a plurality of times t_i^j; ∈ [t_i1, t_im] (j = 1,...n) during inspiration and a plurality of times te^k ∈ [t_e1, t_em] (k = 1,...h) during expiration and are stored as amplitude distributions of the values of the signal for breathing activity at this plurality of times. These amplitude distributions can be used as follows to determine the path of the threshold curve to the target threshold value. For this purpose, advantage is first taken of the fact that the distribution of the phase durations constitutes a probability density function, which can be converted (by integration) into a distribution function V(t) which then increases from 0 at the lowest point of the density function (the shortest phase duration observed) to 1 at the extreme end point of the distribution (the longest phase duration observed). The value of this distribution function at any given time says how probable it is that a changeover to the next phase of ventilation has taken place by the time in question. The probability of a changeover, which increases with time, can be converted into thresholds which go down in a corresponding way in the amplitude distributions of the signal for breathing activity, at the plurality of successive times, in such a way that the probability of a changeover given by the distribution function of the phase durations follows the probability with which, according to the amplitude function, the threshold criterion for changeover will have been met.

[0018] The thresholds in the amplitude distributions may for example be set in such a way that the distribution function V(t) of the phase durations at the plurality of times t_i^j ∈ [t_i1, t_im] and te^k ∈[t_e1, t_em] defines a p quantile criterion in the distributions of the signal for breathing activity, where p is a function p = F(V(t)) of V(t). The function which F(V(t)) is of the distribution in this case is one which generally varies substantially linearly with the distribution function, being in the simplest case the identicality F(V(t)) = V(t) for expiration and the reflection F(V(t)) = 1-V(t) for inspiration.

[0019] Alternatively, the thresholds may be set as quantities A(V(t_ij)) and A(V(t_e^k)) (generally not whole number-quantities) of standard deviations relative to the mean of a Gaussian distribution in such a way that the probability of the phase end given by the distribution of the phase durations is the same as the probability of the threshold criterion being met in the amplitude distributions. A(V(t)) is a function determined in advance which fixes the quantity of standard deviations such that the probability of the phase end given by the distribution of the phase durations is the same as the probability of the threshold criterion being met in the amplitude distributions. This function may for example be so selected that the probability of the phase end given by the distribution function V(t) of the phase durations at a time t follows the probability of the threshold criterion being met in the amplitude distribution, for which purpose it is possible to use the (tabulated) Gaussian error integral


which states the probability with which there is a value above a value u in a Gaussian distribution, e.g. in a Gaussian distribution, 68% of the entries are within 1σ, 95% are within 2σ and 99.7% are within 3σ (σ = standard deviation).

[0020] As an example, the function F(V(T)) = 1 - V(t) is set in inspiration and the function F(V(T)) = V(t) is set in expiration. In the histograms of the amplitude distributions of the signal for breathing activity at the times t_i^j ∈ [t_¡1, t_im] and t_e^k ∈ [t_e1, t_em], the thresholds are then set in such a way that they fix a (1-V(t_i^j))quantile in the amplitude distribution in inspiration and a V(t_e^k)) quantile in expiration. For example, let the first time after t¡¹ of the plurality of times be such that the value of the distribution function V(t¡¹) is then 0.05 (corresponding to 0.5%), which is equivalent to a probability of 5% for an actual phase end. At this time, the threshold is then so set in this example in the associated amplitude distribution for the signal for breathing activity that it forms a (1-0.05) quantile, i.e., that 5% of the amplitudes are above the inspiratory threshold which has been set and 95% are below it. At the next time t_i², let the value of the distribution function of the phase durations then be 0.25. The inspiratory threshold in the amplitude distribution will then be set as a (1-025) or 0.75 quantile at this next time, and 25% of the amplitude values will thus be above the threshold and 75% below it. Hence, the threshold is set in such a way that the probability of a changeover to the next phase of ventilation exactly follows the probability which is found for a phase end from the distribution of the phase durations. This process is continued until such time as the value of the distribution function of the phase durations reaches 0.5, which then corresponds to a threshold in the amplitude distributions at this time the value of whose p quantile is p = 0.5.

[0021] The threshold curve may be interpolated in the intervals between the times making up the plurality of times, and for example the threshold curve may in each case be continued linearly to the threshold value at the next time in the plurality of times.

[0022] The threshold is then lowered to the target threshold value, which can be derived from the amplitude distribution at the time when the value of the distribution function of the phase durations is 0.5, i.e. even in the period after the distribution function has reached the value of 0.5 the distribution at that time is taken as a basis and the threshold is then lowered to the target threshold value in this distribution. As described above, this target threshold value is fixed as a p quantile in the amplitude distribution at the time t_im, e.g. as a 0.05 quantile, which means that 95% of the amplitudes of the signal for breathing activity will be above the target threshold value at the time t_im. The consideration underlying this procedure is that after the mean duration of a phase (corresponding to a value of 0.5 for the distribution function) there are already an increasing number of signals for breathing activity from inspirations which are beginning, and there is thus increasing evidence in the distributions of effects of fresh phases of ventilation which have already recommenced and the distributions thus no longer reflect what happens in phases of ventilation of very long durations which are drawing to a close.

[0023] In this case too the continuation to the target values may take place via a plurality of reference points, i.e. at a plurality of times during the interval [t_i1, t_im] or [t_e1, t_em], as the case may be, the threshold is continued as the (1-(V(t)) quantile in the amplitude distribution for the time t_im in the case of the inspiratory threshold and as the (V(t) quantile in the amplitude distribution for the time t_em in the case of the expiratory threshold. The threshold curve may once again be interpolated between the times making up the plurality of times.

[0024] In this case too the following of a path to the target value may take place via a plurality of reference points, i.e. at a plurality of times from the range [t_i1, t_im] or [t_e1, t_em], the threshold follows a path defined by the (1-V(t)) quantile in the amplitude distribution at the time t_im in the case of the expiratory threshold and one defined by the V(t) in the amplitude distribution at the time t_em in the case of the expiratory threshold. The threshold curve may again be interpolated in the intervals between the times making up the plurality of times.

[0025] The present invention will now be described, by way of example, only, by reference to the accompanying drawings of which:

Fig. 1 is a graph which shows, as a function of time, signals for breathing activity for a plurality of breathing cycles, the signals being superimposed on one another; and

Fig. 2 is a graph which shows the signal for breathing activity as a function of time together with the associated threshold value curves.

[0026] In the present example, what is used as a signal for breathing activity is an electromyographic signal (EMG signal) which is picked up from electrodes on the thorax and which represents the muscle activity connected with breathing.

[0027] The EMG signal used is preferably a pre-processed one. Pre-processing of this kind of the raw EMG signal is performed in a known manner in such a way that the raw EMG signal is freed of interfering and disrupting signals (e.g. ECG, motion artefacts, mains hum), and finally, envelope curve detection is carried out. Envelope curve detection may for example be performed by "rectification" followed by low pass filtering, the "rectification" being performed by an operation which gives a value (e.g. squaring or pure absolute-value generation). After the low pass filtering, what is then obtained is the envelope curve, i.e. the curve which is an envelope enclosing the waveform of the raw signal. A preferred implementation of the envelope curve detection process is the formation of what is referred to as the RMS (root mean square) over the length of a sliding time slot. The concept of EMG amplitude estimation which can be described by the term "envelope curve detection" is described in detail in Merletti R, Parker P.A.: Electromyography, Physiology, Engineering and Noninvasive Applications, IEEE Press, Wiley Interscience, 2004, Chapter 6.4 et seq, i.e. page 139 ff.

[0028] An EMG envelope curve signal of this kind was picked up over a plurality of breathing cycles and, as shown in Fig. 1, was superimposed. When this was done, the individual signals were superimposed in such a way that the exact time of the beginning of inspiration (situated at approx. 1.9 s in Fig. 1) which was found, after the beginning of inspiration, by examining the previous signal for breathing activity was positioned at a common time and, this being the case, the successive phases of ventilation were "synchronised" in the plot in Fig. 1.

[0029] In accordance with the invention, there is used in an expiration phase (declining signal for breathing activity) a dynamic threshold curve for the changeover to the next inspiration phase which, after the beginning of the current expiration phase, is kept at high values, and in this example at a high and constant value, for a selected inspiratory refractory period, i.e. until a time t_i1, to prevent a premature changeover to inspiration. This threshold value curve is identified as 2 in Fig. 1. The inspiratory refractory period is a short period of time which is selected in advance or, as will be explained below, is determined from the signals for breathing activity and which begins at the beginning of the expiration phase and ends at the time t_i1 and which is so short that it is extremely unlikely that a new inspiration will begin so short a time after the beginning of the expiration. When a value selected in advance is used, the inspiratory refractory period may be 200 ms for example. During this period the threshold 2 is held at a value so high that the possibility of a changeover to inspiration can be virtually ruled out.

[0030] After the inspiratory refractory period, the threshold is lowered in such a way that an optimum threshold value is fixed for the triggering of the next correct inspiration. For this purpose, the distribution of the expiratory phase durations, which is indicated as 7 in Fig. 1, is first considered. This distribution 7 constitutes a probability density function which, when integrated, can also be plotted as a distribution function V(t), as shown at the top of Fig. 2. The distribution function corresponding to the probability density function 7 would then rise, in Fig. 1, from the time identified as 8, from a very small value (which is determined by the p quantile which was used as described above for determining the inspiratory and expiratory refractory times) to a value of close to 1 (which is determined by the p quantile which was used as described above for determining the maximum phase durations) at the time identified as 11, i.e. the probability of the next inspiration beginning rises in a corresponding way over this interval of time. The curve of the distribution function V(t) is identified as 12 at the top of Fig. 2.

[0031] In a preferred embodiment, the amplitude distribution of the signal for breathing activity is stored at a plurality of times during the interval between the end t_i1 of the inspiratory refractory period and the time t_im of the mean expected phase duration (the median value of the distribution 7 identified as 10 in Fig. 1); in Fig. 1 these times are, by way of example, three, namely 8, 9 and 10. In the histograms 3, 4 and 5 of the amplitude distributions at these times the threshold is then so set in their respective cases that the threshold for the inspiratory changeover corresponds to a p quantile whose parameter is p = 1-V(t_i^k), where V(t¡^k) are the distribution functions of the phase durations at the times k = 8, 9 and 10 as identified in Fig. 1. In more general terms, the p quantiles may be defined as a function of the probability of the distribution function V(t_¡^k), i.e. p = F(V(t_i^k)).

[0032] Alternatively, a Gaussian amplitude distribution may be assumed and the threshold may be defined as p(t_i^k) + k(V(t_i^k)) * σ(t_i^k), where µ(t_i^k) is the mean value and σ(t_i^k) is the standard deviation of the amplitude distribution at time t_i^k. k(V(t_i^k)) is a factor which depends on the probability of the distribution function V(t_i^k). If the amplitude distribution is confined to an interval between µ +/- 2.5 σ, then k = -2.5 + (1-V(t_i^k)) * 5. For very low probability values V(t_i^k), k = 2.5 and the threshold is thus relatively high for µ(t_i^k) + 2.5 * σ(t_¡^k). At high probabilities (approaching 1), the result is a low threshold for p(t_i^k) - 2.5 * σ(t_i^k).

[0033] Clearly, what this means is that, after beginning, at time t_¡1, at the topmost edge of the family of curves representing the signals for breathing activity, the threshold then, in the course of the transition to time t_im, cuts increasingly deeply into the amplitude distributions of the signals for breathing activity in the histograms 3, 4 and 5, as indicated by the parts of the distributions shown in black in Fig. 1 and Fig. 2, which indicate the amplitudes of the signal for breathing activity which exceed the inspiratory threshold. At the time t_im, the mean duration of the phase, where V(t_im)= 5, the threshold is so positioned that it forms the median or the 0.5 quantile of the amplitude distribution of the signal for breathing activity at time t_im. In line with this, the threshold is then situated in the centre of the distribution 5. Once the mean phase duration t_im has been reached, there is no point in the threshold continuing to be oriented to histograms of the amplitude distributions at later times, because these later histograms would increasingly contain effects of signals for breathing activity attributable to inspirations which have already begun. The threshold is therefore now lowered to the target threshold value until the expected maximum phase duration t_i2, identified as 11 in Fig. 1. As was described above in connection with the times t_i^k, the target threshold value is determined as a quantile or, if there is assumed to be a Gaussian distribution, as a multiple of the standard deviation relative to the mean, but referred to the amplitude distribution at the time t_im.

[0034] During an inspiration phase, i.e. for the detection of the next expiration, the dynamic expiratory threshold follows a path of the type described above except it is held at low values until the end of the expiratory refractory period t_e1 and is then raised in a monotonic increase to an expiratory target threshold value, i.e. the threshold follows the reverse path to that shown in Figs. 1 and 2 and moves from below into the distributions until, at the median value t_eM of the expiration phase durations it is situated in the centre of the distribution of the amplitudes of the signal for breathing activity, after which it is raised in a monotonic increase to the expiratory target threshold value. In the same way as was described above, this target threshold value is determined as a quantile or, if there is assumed to be a Gaussian distribution, as a multiple of the standard deviation relative to the mean, but referred to the amplitude distribution at the time t_em.

Claims

1. A method for the automatic control of a ventilator machine, for changing over between two alternating phases of ventilation (inspiration and expiration) by, in one phase of ventilation, causing a control unit to examine a sensed respiratory signal for breathing activity for a threshold criterion for the changeover to the next phase of ventilation and changing over from one phase of ventilation to the other when the threshold criterion is met, wherein there is used in an expiration phase, for the changeover to an inspiration phase, a dynamic threshold curve U_insp,_thresh(t) which, after the beginning of the current expiration phase is kept at high values until the end t_i1 of a selected inspiratory refractory period to prevent a changeover to inspiration and thereafter is lowered in a monotonic decrease to an inspiratory target threshold value at the expected time t_i2 of the maximum duration of the current expiration phase, and a changeover is made to the inspiration phase when the signal for breathing activity rises above the threshold curve for inspiration, and in an inspiration phase, there is used for the changeover to an expiration phase a dynamic threshold curve U_exp._thresh(t) which, after the beginning of the current inspiration phase, is held at low values until the end t_e1 of a selected expiratory refractory period to prevent a changeover to expiration and thereafter is raised in a monotonic increase to an expiratory target threshold value at the expected time t_e2 of the maximum duration of the current inspiration phase, and a changeover is made to the expiration phase when the signal for breathing activity drops below the threshold curve for expiration, and the durations of the inspiration and expiration phases or the duration of a breath (an inspiration and an expiration phase) are each stored and the expected times t_i2 and t_e2 of the maximum phase durations are derived from the distributions of the phase durations when the reference point is the beginning of the given phase of ventilation, or from the distribution of the breath durations when the reference point is the beginning of the previous phase of ventilation.

2. The method according to claim 1, in which the expected maximum phase durations t_i2 and t_e2 are derived from the distributions of the durations of the inspiration and expiration phases or from the distribution of the breath durations, in the form of a P quantile of these distributions, the parameters P_i2 and P_e2 for the quantiles being fixed in advance and the values of these parameters P_i2 and P_e2 being at least 0.8.

3. The method according to claim 1, in which the expected maximum phase durations t_i2 and t_e2 are derived from the distributions of the durations of the inspiration or expiration phases or from the distribution of the breath durations on the assumption that the distributions follow a Gaussian distribution, and the expected maximum phase durations are fixed by a quantity of standard deviations above the mean value which correspond to a preset probability that a maximum phase duration will be situated before the fixed end point of a phase, this probability being at least 0.8

4. The method according to one of the preceding claims in which the times t_i1 and t_e1 of the ends of the inspiratory and expiratory refractory periods and the expected times t_i2 and t_e2 of the maximum phase durations are referred to the time of the changeover to the current phase of ventilation.

5. The method according to claim 1, 2 or 3, in which the waveform of the signal for breathing activity is stored over at least the duration of one half-phase of ventilation and the beginning of the current phase of ventilation is determined by examining the waveform of the signal for breathing activity in a period around the time of the changeover to the current phase of ventilation and the times t_¡1 and t_e2 of the ends of the inspiratory and expiratory refractory periods and the expected times t_i2 and t_e2 of the maximum phase durations are referred to the time of the beginning of the current phase of ventilation which is determined in this way.

6. The method according to one of the preceding claims, in which the times t_i1 and/or t_e1 of the ends of the inspiratory and/or expiratory refractory periods are derived from the distributions of the durations of the inspiration and expiration phases as respective p quantiles of the said distributions, the p quantile values p_i1 and P_e1 which are used in this case being selected in advance and their values being less than 0.1.

7. The method according to one of the preceding claims, in which the values of the signal for breathing activity are stored at a plurality of times t_i^j ∈ [t_i1, t_im] and t_ek ∈ [t_e1, t_em] (i = 1,...n and k = 1,...m) where t_im and t_em are the times of the median values of the distributions of the inspiratory and expiratory phase durations, and are amassed as amplitude distributions of the values of the signal for breathing activity at this plurality of times, and the inspiratory and expiratory thresholds are so set in these amplitude distributions that the probability of changeover given by the distribution function of the phase durations at the times t_i^j ∈ [t_¡1, t_im] and t_e^k ∈ [t_e1, t_em] follows the probability with which, according to the amplitude distributions at the times t_i^j and t_e^k, the threshold criterion for changeover will be met.

8. The method according to claim 7, characterised in that the inspiratory threshold U_insp,_thresh(t_i^j) is fixed in such a way that it forms a p quantile where p = F1(V(t_i^j)) in the distribution of the signal for breathing activity at the time t_i^j, where F₁(V(t_¡^j)) is a function of the distribution function of the phase durations which is determined in advance, after which the inspiratory threshold is lowered in such a way that it reaches the inspiratory target threshold value at the expected maximum phase duration t_i2, and the expiratory threshold U_exp,_thresh(t_e^k) is fixed in such a way that it forms a p quantile where p = F₂(V(t_e^k)) in the distribution of the signal for breathing activity at the time t_e^k, where F₂(V(t_e^k)) is a function of the distribution function of the phase durations which is determined in advance, after which the expiratory threshold is raised in such a way that it reaches the expiratory target threshold value at the expected maximum phase duration t_e2.

9. The method according to claim 8, characterised in that F₁(V(t)) = 1-V(t) and F₂(V(t)) = V(t) are defined in advance.

10. The method according to claim 7, characterised in that, on the assumption of a Gaussian amplitude distribution, the inspiratory threshold is defined as µ(t_i^j) + k(V(t_i^j)) * σ(t_i^j), where µ(t_i^j) is the mean value and σ(t_i^j) is the standard deviation of the amplitude distribution at the time t_i^j and k(V(t_i^j)) is a factor which has a dependence which is selected in advance on the distribution function V(t_i^j), after which the inspiratory threshold is lowered in such a way that it reaches the inspiratory target threshold value at the expected maximum phase duration t_i2, and in that, on the assumption of Gaussian amplitude distributions, the expiratory threshold is defined as p(t_e^k) + k(V(t_e^k)) * σ(t_e^k), where µ(t_e^k) is the mean value and σ(t_e^k) is the standard deviation of the amplitude distribution at the time t_e^k and k(V(t_e^k)) is a factor which has a dependence which is selected in advance on the distribution function V(t_i^k) , after which the expiratory threshold is raised in such a way that it reaches the expiratory target threshold value at the expected maximum phase duration t_e2.

11. The method according to one of claims 7 to 10, characterised in that the curves of the inspiratory and expiratory thresholds are interpolated between successive times t_i^j and t_e^k respectively.

12. A ventilator machine having a source of gas for breathing, a ventilator unit for feeding gas for breathing through connecting lines, a control unit which controls the ventilator unit, sensors for picking up at least one pneumatic signal for breathing activity which are connected to the control unit, characterised in that the control unit is arranged to carry out a method according to one of the preceding claims.

Drawing